This story appeared on Network World at


Key decisions that enabled the 'Net


'Net Insider 


By Scott Bradner, Network World, 10/02/06


From time to time it is useful to look back to see how one got to someplace. I was reminded that this might be a good time to look back to see how we got to the Internet recently when I got an e-mail from Dennis Jennings noting that the NSFNet project was approved 21 years ago.


By one measure the Internet has reached the age of majority in Washington, D.C. - one of the places that seems to have the worst understanding of what the Internet was, is and can be. The technology trickle that became the Internet started with research into packet-based networks in the early to mid 1960s by Len Kleinrock, Larry Roberts, Paul Baron and others (Google can help you find lots of information on these folks).


This research led to a 1967 design session run by Roberts that in turn led to a request for proposals for what was then called an IMP but is now called a router. BBN won the RFP, and the first four IMPs were installed in 1969, creating the start of the ARPANET. By 1971, 15 IMPs, including one at Harvard University, were installed and interconnected, and the ARPANET quickly became the important way for federally funded researchers to communicate.


But access to the ARPANET was limited to those researchers and staff on the connected machines. I've come up with a list of 10 key decisions that got us from those early days to the Internet we have today (for a sidebar, see Internet histories compiled by the Internet Society. They are:


  Using existing networks instead of creating an entirely new infrastructure.


  Using packets rather than circuits, so there is no reason for a carrier to be involved in setting up communications.


  Creating router function to logically isolate sections of the network.


  Splitting TCP and IP and making the level of reliability an end system option.


  U.S. government funds UC Berkeley to add TCP/IP to Unix and make software easily available.


  Computer Science Network sites use ARPANET for e-mail, which enables anyone at the university to use the ARPANET in 1981, the start of generations of students who regularly used e-mail.


  Dennis Jennings, manager of NSFNet, requires the use of TCP/IP on NSFNet, reinforcing TCP/IP as the standard.


  The International Standards Organization turns down an offer to take over TCP/IP standardization - if it had accepted, the Internet would be carriercentric rather than open.


  NSF blocks the use of NSFNet for commercial use, which forced the development of commercial ISPs.


  U.S. government imposes no significant regulations on the Internet, letting innovation run free.


Note that these decisions, except that of Jennings about TCP/IP, facilitated rather than mandated actions. That attitude is well behind us in Washington.


The feeling there now is that the Internet is far too regulation-free (and maybe too good at innovating). Congress and the FCC are fighting to fix this perceived problem.


I have no idea if we will be able to look at the Internet 10 years from now and see anything we would recognize as the Internet. Many in Washington seem to hope not.


Disclaimer: I predict that Harvard will look more like Harvard in 10 years (one-37th of its age) than the Internet will look like the Internet in 10 years (one-fourth of its age), but that is my prediction, not Harvard's.


All contents copyright 1995-2006 Network World, Inc.